Pareto Partial Dominance MOEA and Hybrid Archiving Strategy Included CDAS in Many-Objective Optimization


In this work, we propose a novel multi-objective evolutionary algorithm (MOEA) that uses Pareto partial dominance, which calculates dominance between solutions using only r objective functions selected from m objective functions to induce appropriate selection pressure in the evolution process of MOEA. Also, we temporally switch r objective functions among C-m(r) combinations in every interval generations I-g to optimize all of the objective functions throughout the entire evolution process. In this work, we use many-objective 0/1 knapsack problems to verify the search performance of the proposed Pareto partial dominance MOEA (PPD-MOEA). Simulation results show that there is an optimum value for the number of objective functions r to be considered in Pareto partial dominance, and the interval (generation numbers) I-g to maximize the entire search performance. Also, the search performance of PPD-MOEA is superior to NSGA-II and recent state-of-the-art MOEAs, i.e., IBEA, CDAS and MSOPS. Additionally, to further enhance the search performance of PPD-MOEA, we propose a hybrid archiving strategy which uses both conventional NSGA-II and CDAS to select well-spread and well-converged solutions simultaneously when updating the archive population. Simulation results show that the hybrid archiving strategy further improves the search performance of PPD-MOEA by enhancing convergence while maintaining diversity in the archive population.